Académique Documents
Professionnel Documents
Culture Documents
North-Holland
evidence*
This paper provides evidence on the cross-sectional relation between firms investment opportunities, their debt and compensation contracts. their size and financial leverage. and their accounting
procedure choices. This evidence is important. because previous studies hypcthesize !!?ut the link
between firms investment oppor!-unities and their accounting choices helps explain extant results on
the size. debt/equity. and bonus plan hypotheses. However. while 1 find that firms rnvestment
opportunities do affect the nature of their coIltracts, I also find that the traditionalexplanations for
accounting choice are important after controlling for the effects of the investment opportunity set.
1. Introduction
01654101/93~$06.00
( 1993
408
D. J. Skinner,
Invesrmenr
opportuniries
and accounting
choice
409
2Along similar lines, Ball and Foster (1982) and Holthausen and Leftwich (1983) had originally
criticized these types of accounting choice studies because of the likelihc: ,d that the size, debt/equity,
and bonus plan proxy variables that they employed are likely to be correlated with many other
things, including industry membership.
410
nonetheless increases our understanding of extant accounting practice, particularly when combined with the evidence from studies that examine smaller
samples of firms in particular economic circumstances (such as Zimmers).
The next section of the paper summarizes existing theory on the relation
between the ios and accounting choice, and emphasizes that it is not obvious
a priori how the ios affects the accepted set of accounting procedures. In
addition, I provide a rationale for why we expect a relation between the ios and
the nature of firm contracts, and so why there is also likely to be an indirect link
between the ios and accounting choice. Section 3 provides details of the sample.
Section 4 discusses the proxy variables that I use to measure the ios, and
documents how the ios covaries with firm size, financial leverage, and long-term
financial performance. Section 5 provides evidence of how the ios affects the
nature of firms bonus plans and debt contracts. Section 6 then provides
evidence of variation in firms accounting choices, and documents how this
variation is related to variation in these firms size, financial leverage, debt and
compensation contracts, and investment opportunities. Section 7 provides a discussion and summary.
In
ios
ios
to
so
Watts and Zimmerman (1986, 1990j isolate two principal forces that determine firms accounting procedure choices. First, certain accounting procedures
(the accepted set) evolve through time and emerge as best practice - these are
the accounting procedures that cost-effectively resolve the firms agency problems ex ante. This perspective on accounting choice is also known as the
efficient contracting perspective [e.g., Holthausen ( 1990)].3 Second, because it
3An early example of this argument is Watts (1977, p. 61) who draws a linkage between the agency
costs of debt, the nature of firms assets, and the requirement that managers set aside profits for
renewal, repairs, maintenance or depreciation of existing assets. Leftwich (1983) argues that the
accounting choices that have evolved as best practice in private lending agreements (as exemplified
by their existence as boiler-plate type provisions) are those procedures that minimize the costs of
the agency relationship between the firm and its private lenders. Zimmer (1986) provides an example
in the context of Australian real estate developers of how accounting procedure choices can result
from contracting solutions that are rx ~tte efficient.
D.J. Skinner,
Inwstment
opportunities
and accounring
choice
411
4Choices that are made after contracts are in place are not always the result of managerial
opportunism - managers also make choices that are KYpost efficient from the point of view of the
firms claimants.
In addition to Watts and Zimmerman this general point has previously been made by Demski,
Patell, and Wolfson (1984) who argue that the delegation to managers of the choice from among the
set of acceptable (accounting) alternatives can best be understood as efficient, equilibrium behavior
(p. 17) and that by both allowing mangers freedom of choice with respect to accounting methods
and compensating t/tern according/y, owners can capitalize upon managers local expertise (p. 31.
emphasis in original).
412
the accounting choices that are optimal for growth firms probably depend on
as-yet unobservable outcomes, so it may be more costly to restrict accounting
choices in the current period for these firms.)
While it is clear that the ios will affect both the costs and benefits of writing,
monitoring, and enforcing contracts, it is difficult to arrive at more specific
predictions about how exactly the ios will affect the accepted set of accounting
procedures. As Watts and Zimmerman (1986, p. 359) note: Development of
a complex model of the accepted procedure set is difficult since it would involve
modeling the demand for contracting as a monitoring device and the cost of
alternative methods of monitoring. As a result, I adopt the more modest
objective of providing descriptive evidence on the relation between the ios and
observed accounting choicle.
2.2. The indirect relation between the ios and accounting choice
In addition to its direct eXects on the accepted set, the ios is likely to affect the
types of contracts that the firm writes with various claimants. To illustrate this
in a concrete way, I next analyze how the ios affects the nature of management
compensation hnd debt contracts, respectively.
2.2. I. Management compensation agreements
Management compensation agreements help reduce the conflict of interest
between corporate managers and stockholders; these plans are designed to
motivate managers to maximize firm value [Smith and Watts (1982)]. Because
the costs and benefits of monitoring and motivating managers depend on the
nature of the firms investment opportunities, the structure of management
compensation agreements will vary across firms as a function of the ios. Smith
and Watts (1991) argue that the actions of managers of firms with relatively
more assets-in-place are less costly to monitor than the actions of managers of
firms whose value is comprised largely of growth opportunities. In addition,
managers of firms with relatively more growth opportunities are likely to be
allowed more decision-making discretion because these managers have better
information about the firms investment opportunities than the firms stockholders; i.e., managers of growth firms ari: likely to have relatively more specific
knowledge than managers of firms with relatively more assets-in-place. Consequently, Smith and Watts predict that growth firms are more likely to use
incentive compensation schemes that tie management compensation to
measures of firm performance (such as accounting earnings or stock price).
In addition, Smith and Watts (1991) argue that accounting numbers are
poorer performance measures for firms with relatively more growth opportunities because of conservatism in accounting: the need for objective and verifiable
numbers limits the extent to which accountants are willing to recognize income
that depends on uncertain future events. [Consistent with this, Collins et al.
413
(1992) provide evidence that stock prices lead accounting earnings to a greater
extent for firms with relatively large amounts of new and intangible assets.] This
countervailing measurement effect implies that incentive compensation schemes
that rely on accounting performance measures are less likely to be used in firms
with growth opportunities, offsetting the arguments above. Stock-based incentive plans are likely to be more prevalent in growth firms, since these plans
achieve the incentive-alignment benefits without incurring the accounting
measurement costs.6
This discussion is summarized in fig. 1. The figure illustrates the two countervailing effects of the ios on the nature of managerial compensation - firms
with more assets-in-place are: (1) less likely to use incentive compensation
(because their managers optimally have less decision-making discretion) and (2)
more likely to use accounting earnings numbers in compensation contracts
(because these numbers are relatively better performance measures in these
firms).
Smith and Watts (1991) and Gaver and Gaver (1993) provide evidenc+cthat is
consistent with this. Smith and Watts present industry-level evidence that firms
with growth opportunities are (i) less likely to use accounting-based bonus plans
than other firms and (ii) more likely to use stock-option plans. Gaver and Gaver
(1993) replicate the Smith and Watts analysis for 237 growth firms and 237
nongrowth firms, and find that their growth firms are more likely to use
stock-option plans, but no more likely to use earnings-based bonus plans, than
their nongrowth firms. The fact that the evidence indicates that growth firm: are
more likely to use stock-option plans, but is inconclusive with respect to which
firms are more likely to use bonus plans, is consistent with the existence of these
two opposing effects.
Overall, it is difficult to make a clear a priori prediction about the relation
between the ios and the use of accounting-based bonus plans because of the two
countervailing effects, However, the evidence in Smith and Watts (1991) and the
evidence below (table 3) suggests that, at least as an empirical matter, firms with
relatively more assets-in-place are more likely to use bonus plans that rely on
accounting numbers. Consequently, in fig. 1 I indicate that firms with more
assets-in-place are more likely to employ earnings-based bonus plans. If this is in
Sibson (1990, pp. 342-343) prescribes for fast-growing enterprises: There is continuous change
in these firms. Therefore, compensation systems must be simple and highly adaptable to
change . . . In fast-growing companies, you dont have annual bonus plans. Performance standards
would be changing frequently. Also, some plans, like management bonus plans, can be divisive in
a fast-growth environment.
7See also Clinch (1991) who finds that high-R&D firms are more likely to use stock-option plans
than low-R&D firms, but that there is no apparent difference in these firms propensity to use
earnings-based bonus plans, and Lambert and Larcker (1987). who find that firms with relatively
high rates of sales and asset growth place more weight on market returns and less weight on
accounting numbers in determining managerial compensation than do other firms.
414
D.J. Skinner,
Investment
opportunities
and accounting
choice
Less likely to
use incentive
compensation
__t
L
Accounting numbers
are relatively good
measuresof
performance
I
Opportunism Linkage
Fig. 1. The link between the investment opportunity set, incentive compensation, accounting
earnings-based bonus plans, and accounting procedure choices. (*The fact that firms with more
assets-in-place are more likely to use earnings-based bonus plans is an empirical statement based on
the evidence in this and o:her studies.)
fact the case in general, and if the contractual terms of bonus plans provide
managers with incentives to make income-increasing
accounting procedure
choices, then I expect that firms with more assets-in-place are more likely to use
accounting-based bonus plans, so that their managers are more likely to select
income-increasing accounting procedures ex post. This is shown as the last
linkage in fig. 1.
Debt convers-ts
are written to reduce the conflict of interest between the
firms stockholders and bondholders [Smith and Warner (1979)]. Smith and
Warner describe four principal sources of conflict between bondholders and
stockholders, two of which have implications for the ios-accounting
choice
linkage.
*Unlike the accruals choices that Healy (1985) documents, managers accounting procedure
choices cannot be altered every period depending on where the firms accounting income falls
relative to the upper and lower bounds in bonus plan contracts; it is relatively costly to change
accounting procedures [see e.g., Sweeney (199l)J Therefore, the relation between the existence of
bonus plans and managers* accounting procedure choices is likely to depend on whether incomeincreasing procedure choices will, on average over a period of years, increase the managers bonus.
Based on the relatively strong evidence in favor of the bonus plan hypothesis, this is probably the
case. However, this is ultimately an empirical question and in the tests that follow I provide evidence
about the relation between the ios and the terms of bonus plans for a subsample of firms with
available data.
The other two sources of stockholder-bondholder
conflict are dividend payment (under which
the stockholders opportunistically reduce the value of the bondholders claims by reducing the firms
level of investment to increase the dividend) and claim dilution (under which stockholders
opportunistically dilute the value ofexisting bondholdersclaims by issuing more debt of the same or
higher priority). Neither of these conflicts relates directly to the nature of the firms assets, and so
neither of these conflicts is linked to the firms ios.
415
Press and Weintrop (1990) provide evidence that firms that use accounting-based debt covenants are more highly levered than firms that do not have these constraints. If firms with more
assets-in-place are also more highly levered, this evidence supports the view that the ios affects the
likelihood that firms will have accounting-based debt covenants.
416
Higher financial
leverage
Accounting numbers
are relatively good
measuresof
performance
*I
OpportunismLinkage
Fig. 2. The link between the investment opportunity set, financial leverage, accounting-based debt
covenants, and accounting procedure choices.
In section 2.2.1, the opportunism to which I was referring involved wealth transfers from the
firm to its managers under managerial compensation contracts. In this section, I am referring to
opportunism under debt agreements that involves wealth transfers from bondholders to stockholders that are implemented by managers (who are, following Smith and Warner, assumed to act in the
best ir,:eres& of the stockholders). In both cases managers choose income-increasing accounting
procedures .I effect the wealth trpn$ers.
D. J. Skinner,
hcestnzennr opporruniries
und accounting
choice
417
between the firm and its stakeholders, and because - other things equal - the
nature of these contracts, in turn, affects the accounting procedures that managers select ex post. The ceteris paribus assumption here is important, because
the effects of the ios on the accepted set of accounting procedures may offset
managers opportunistic incentives (this is an empirical question). For example,
if firms with relatively more assets-in-place are more likely to use earnings-based
bonus plans, so that managers of these firms have larger incentives to make
income-increasing accounting choices cx post, it may also be that these firms
accepted sets of accounting procedures are more likely to restrict managers
ability to make these choices ex ante.
3. Sample selection
To obtain
a broad cross-section
418
D.J. Skinner,
Inoestmenr
opportunities
and accounting
choice
Table 1
Summary of SIC-defined industry representation of 504 sample firms with available data in 1987.
Industry description
SIC codes
Number of
sample firmsa
Heavy Construction
Food Products
Textile Mills
Finished Apparel
Millwork, etc.
Furniture
Paper Products
Newspaper Publishing
Periodicals, Books
Chemicals
Drugs, Pharmaceuticals
Plastics
Footwear
Cement
Blast Furnaces (Steel)
Nonferrous Metals
Metal Tools, Parts, etc.
Machinery
Computing Equipment
Electrical Machinery
Semiconductors
Motor Vehicles
Aircraft/Aerospace
Instruments & Controls
Air Transport
Radio & TV Broadcasting
Retail - Dept. Stores
Retail - Grocery
Retail - Apparel
Retail - Fast Food
1600- 1799
2000-2070
2200-222 1
2300-2330
2421-2451
2510-2522
2600-263 1
2711
2721-2731
2800-2829
2834
3080- 3089
3140
3241-3270
3312
3330-3334
341 f-3460
3510-3537
3570-3571
3600-362 I
3674
3711
3721-3760
3812-3861
4512
4833
531 l-5331
5411
5651
5812
10
34
13
18
11
12
22
11
6
24
27
15
10
11
14
11
33
20
19
14
14
7
28
51
10
6
20
9
6
18
12
47
17
25
14
15
23
13
6
33
39
18
10
II
23
14
40
?2
23
19
15
10
32
75
15
7
26
13
6
27
504
650
Total firms
Population
of firmsb
The number of firms in a given industry group in the final sample. This represents those firms in
the population (see footnote b below) that are incorporated in the United States, have 1987 annual
reports on tile at the Universi!y of Michigan Kresge Business Administration Library, and that
disclose their accounting choices.
bThe populatio n represents all firms in a given industry group on the 1989 annual Compustot
industria! tape.
The accounting procedure choices that I seek to explain are those reported at
fiscal year-end 1987. Therefore, the financial variables are measured for each
firm as the average of the year-end values for 1985 through 1987. The data are
419
The construct that I seek to measure here is Myers (1977) notion of assetsin-place vs. growth opportunities. I use three variables to capture this construct.
In addition to these three variables, I include a measure of the riskiness of these
firms assets.
First, to measure assets-in-place (i.e., those assets whose ultimate value does
not principally depend on future discretionary investment by managers), I calculate the ratio of the book value of gross property, plant, and equipment (PP&E)
to the value of the firm. The intuition behind this measure is that past investments in property, plant, and equipment can be characterized as assets-in-place.
This measure is increasing in the proportion of assets-in-place to value.
Second, as an additional proxy for growth opportunities vs. assets-in-place,
I use the firms research and development (R&D) expense deflated by net sales.
The rationale here is that investments in R&D yield expected payoffs that form
part of managers private information and that the value of these investments is
difficult for outsiders to measure reliably. Moreover, R&D expenditures comprise, at least to some extent, discretionary expenditures and so are like the
growth options that Myers describes [Dechow and Sloan (1991)]. Long and
Malitz (1985) and Smith and Watts (1991) also use R&D expenditures to
measure Myers notion of growth opportunities.
Finally, I also include Tobins 4 ratio, which is defined as the market value of
the firm divided by the replacement cost of its assets. This variable increases with
the proportion of firm value represented by intangible assets. Both Lindenberg
and Ross (1981) and Merck, Shleifer, and Vishny (1988) provide arguments
and/or evidence that Tobins 4 increases with the proportion of value represented by intangibles [in fact Merck, Shleifer, and Vishny (1988, p. 299) use
R&D and advertising variables to control for observable measures of intangible
assets that affect $1.
Following Lindenberg and Ross (1981), I estimate the replacement cost of
these firms assets as the sum of the replacement cost of property, plant, and
equipment (PP&E), the replacement cost of inventories (measured as the book
value of inventories plus the LIFO reserve), and the book value of other assets.
I estimate the replacement cost of PP&E using the procedure outlined by Hall et
al. (1988) for the NBER R&D master file data:
1. Estimate the average age of the PP&E by dividing 1987 year-end accumulated depreciation by the 1987 depreciation expense (this assumes straight-line
depreciation).
2. Multiply the 1987 net book value of PP&E by ihc ratio of the GNP
deflator for fixed nonresidential investment in 1987 to this same deflator AA
420
D.J. Skinner,
Inaesrtnent
opportunities
and accounting
choice
years before 1987, where AA is the average age of the assets from 1. [This
assumes that the replacement cost of PP&E grows, on average, at the same rate
as the GNP deflator for fixed nonresidential investment. Hall et al. (1988) and
Lindenberg and Ross (1981) use the GNP deflator for fixed nonresidential
investment in similar calculations.]
I then divide the market value of the firm (i.e., the size measure described
below) by this estimate of the replacement cost of the assets to yield an estimate
of q.
To proxy for the risk of these firms investment opportunities, I estimate their
asset betas. To calculate an asset beta, I first estimate the beta of each firms
stock, and then unlever this number by multiplying by the equity-to-value ratio
discussed below [see Hamada (1972)]. This estimation procedure assumes that
the firms debt is approximately risk-free.14
To summarize, I use four variables to characterize the ios for these firms: one
measure that is positively related to assets-in-place/firm value (gross PP&Eto-value), two measures that are positively related to growth opportunities/firm
value (R&D-to-sales and q), and a measure of risk (asset beta).
4.1.2. Firm size
and
financial leverage
I measure the size of the firm as the sum of the market value of equity and the
book value of debt (current liabilities plus long-term debt). Two measures of
financial leverage are employed, one based on book values and the other on
market values. The book value measure is the ratio of the book value of
long-term debt to the book value of total assets (debt-to-assets), while the
market value measure is the ratio of the market value of equity to firm size,
calculated as described above (equity-to-value). The debt-to-assets measure is
increasing and the equity-to-value measure is decreasing in financial leverage.
4.1.3. Accounting ROA
Firms past accounting performance is used since there is reason to believe
that accounting procedure choice is related to how well or badly firms are
performing: firms that are poor performers are more likely to select incomeincreasing accounting procedures [see, for example, Lilien, Mellman, and Pastena (1988) and Sweeney (1991)]. It is important to control for accounting
14The betas are estimated for most (305) firms using five years of monthly CRSP data through
December 1987. Some firms are not available on the CRSP monthly returns files, however, and so
for these firms (134) I use 500 days of daily return data from the CRSP daily returns file. I thus have
beta estimates for a total of 439 sample firms. I use montly beta estimates where available because
these are less susceptible to the measurement error that arises because of nonsynchronous
trading
than are daily beta estimates.
D.J. Skinner,
Inresmenl
opportunities
and amounting
choice
421
in addition
to these other
D.J.
422
Skinner,
Incestment
opportunities
und uccounting
choice
Table 2
Sp;arman
rank
_~~~_~~
correlations
for fiscal 1987 between the size, profitability,
financial
investment opportunity
set measures of 504 sample firms.. b
_~~~
- ~~~~~~
-~.-~ ~- --Financial leverage
_..._~~_~_ _~ .-~~
Accg
Size:
Value
opportunity
PPE/
value
Asset
beta
Debt/
assets
0.25
- 0.02
0.17
0.41
0.01
- 0.02
0.21
0.14
0.25
- 0.63
- 0.32
0.62
Financial
lererage:
Debt/assets
MVE,value
and
set variables
~~~~~~_~
ROA
Projitahiliry
Accg ROA
lnresrment
opportunit_v
Asset beta
PPE/value
R&D/sales
MVE,
value
Investment
leverage,
R&D/
sales
0.19
~_~~~~
Tobins
4
0.18
- 0.1 Id
- 0.04
0.21
- 0.38
- 0.24
0.22
- 0.19
0.56
- 0.24
0.29
- O.lOd
0.40
- 0.61
0.22
cariohles:
different
different
test.
test.
opportunities. Finally, y is positively related to the R&D (0.22) measure, consistent with q measuring firms intangible investment opportunities.16
There is also evidence of an association between these firms investment
opportunities and their financial leverage. In particular, financial leverage is
1 also performed a factor analysis to try and gauge how many underlying or latent variables the
four ios variables were capturing. The factor analysis reveals that a single factor explains 49% of the
cross-sectional
variation
in these four variables, and that two factors together explain 73% of the
variation. Moreover,
when 1 cxcludc the risk measure, a single factor explains 55% of the variation
in the remaining
three variables. This evidence is consistent with these variables capturing
two
underlying investmcnl opportunity
constructs: the assets-in-place vs. growth options dimension, as
well as the fact that asset risk varies as a function of the ios.
5. The relation between the investment opportunity set and firm contracts
This section provides evidence on the relation between the ios and nature of
sample fi*?ms management compensation plans and debt contracts. This linkage
is important because the arguments in section 2 suggest that the ios affects
accounting choice indirectly through its effects on the nature of firms contracts.
I collect information
f 1985). and
424
statements that these firms filed over the 1981 through 1990 period. (By selecting
only the largest firms I weaken the power of these tests if, as the correlations in
table 2 suggest, size is correlated with the ios: by choosing large firms I tend to
select firms with relatively more assets-in-place than characterizes the sample as
a whole.) I chose the ten-year sampling interval because previous studies report
that bonus plans are subject to shareholder approval every three, five, or ten
years, at which time details of the plans are usually available in the proxy
statement [e.g., see Healy (1985)].
Of these 100 firms, 9 appear not to have a bonus plan, 6 disclose the existence
of a plan but provide no information about how awards are determined under
the plan, 44 have a discretionary bonus plan (under which the bonus award is
made at the discretion of the compensation committee of the board of directors),
and 41 have incentive bonus plans that explicitly tie the bonus to an accounting
earnings measure (return on equity, operating earnings, net income before taxes,
extraordinary items, etc.).l9 These proportions are comparable to those in Healy
(1985). Healy starts with the Fortune 250, and finds that around half of these
firms (123) have bonus plans but do not publicly disclose bonus plan details, and
ultimately arrives at a sample of 94 firms with bonus plans that tie the bonus to
accounting earnings. These proportions are similar to mine because a large
number of firms in the former group probably have discretionary bonus plans
that Healy excludes.
To investigate whether there is a relation between the ios and the nature of
firms bonus plans, table 3 compares the size, leverage, profitability, and ios
measures for those firms: (i) with incentive bonus plans (41 firms), (ii) with
discretionary bonus plans (44 firms), and (iii) without any bonus plans (9 firms).
If the arguments in section 2 are correct, accounting earnings should be
relatively better performance measures in firms with more assets-in-place.
The terms incentive and discretionary bonus plans are from Sibson (1990, ch. 18). Although
a detailed examination of this issue is best left for further research, my reading of the available details
of these 85 bonus plans indicates clearly that two distinct types of plan exist, where one type
(discretionary) ostensibly provides boards with much more discretion than the other. More
specifically, incentive bonus plans typically specify - with accounting e;mings numbers - the upper
and lower bounds, and then give the compensation committee discretion to decide on the bonus
within these bounds. For example, the General Motors Bonus Plan provides for a transfer to the
reserve of 8% of net earnings in excess of $I billion, provided that amount does not exceed the total
amount of the common stock dividend for the year. In contrast, discretionary plans provide the
compensation committee with almost complete discretion with respect to the bonus. For example,
Union Carbides 1989 proxy statement, in describing the bonus plan, states that the bonuses, if any,
for 1988 will be determined by the Committee. . . on the basis of the Committees evaluation of
factors such as the Corporations financial results . . . , the performance of the officers individually
and . . . as a group, and the levels of salaries and bonuses paid by competitive employers. The
Committee has the right, in its sole discretion, to increase, decrease, or eliminate the bonus
the basis of the Committees review (Union Carbide proxy statement dated March 17. 1988,~.2~
The existence of such a dichotomy makes sense if accounting earnings are better measures of firm
performance in some firms than they are in others, as 1 argue below.
425
Table 3
Comparison of average (median) size, leverage, and investment opportunity set measures for 94
sample firms: (i) with an accouniing-earnings-based incentive bonus plan. (ii) with a discretionary
bonus plan that is not explicitly tied to accounting earnings, and (iii) without any type of bonus plan,
in fiscal 1987.
Firms with a
discretionary
bonus plan
(44)
Firms with an
incentive
bonus plan
(41)
11,747
(11,187)
7,407
(6,638)
14,184
(6,791)
Debt/assets
0.148
(0.106)
0.177
(0.143)
0.207
(0.213)
MVE/value
0.717
(0.809)
0.648
(0.680)
0.608
(0.602)
0.115
(0.109)b
0.124
(0.119)
0.131
(0.134)
Asset beta
0.824b
(0.862)
0.706
(0.705)
0.679
(0.670)
PPE/value
0.474
(0.266)
0.394
(0.344)
0.526
(0.461)
R&D/sales
0.054
(0.075)b
0.042b
(0.023)b
0.026
(0.012)
Tobins q
1.51
(1.52)
1.36
(1.20)
1.25
(1.16)
Firms without
a bonus plan
(9)
Firm size:
Value
Financial leverage:
Projtability:
Accounting ROA
Investment opportunity variables:
The variables are defined as follows. All variables (except asset beta, accounting ROA, and
Tobins q) are calculated for each firm as the average over year-end 1985 through year-end 1987, and
so are based on three observations for each firm.
Value
= market value of the firm (in millions of dollars)
= market value of equity plus book value of debt;
Debt/assets = book-value of long-term debt divided by the book value of total assets;
MVE/value = market value of equity divided by the market value of the firm;
Asset beta = beta of the firms stock multiplied by the MVE/value ratio;
PPE/value
= gross property, plant, and equipment (at historic cost) divided by the market value
of the firm;
R&D/sales = research and development expense divided by net sales, expressed as a percentage;
Tobins q
= estimate of Tobins q ratio (see text for details of calculations).
bStatistically significant difference between the number in this column and the corresponding
number for firms with an incentive bonus plan at the 10% level under two-tailed, two-sample f-tests
(Wilcoxon tests).
Statistically significant difference between the number in this column and the corresponding
number for firms with an incentive bonus plan at the 5% level under two-tailed, two-sample r-tests
(Wilcoxon tests).
426
Consequently, these firms are likely to use incentive bonus plans that tie the
bonus directly to accounting earnings while firms with more growth opportunities are likely to use discretionary bonus plans that do not.
The evidence in table 3 is consistent with this. Firms that use incentive
bonus plans have higher mean and median gross PP&E to value ratios
(more assets-in-place) but smaller q and R&D ratios (fewer growth opportunities) than firms that use discretionary bonus plans, although only the difference in mean and median R&D ratios is statistically significant. Moreover,
firms without any bonus plans appear to be those with the largest proportion
of growth opportunities of all of the three groups, which is also consistent
with accounting numbers being poorer performance measures in growth
firms.
The evidence in table 3 provides some support for the idea that the ios affects
the firms choice of bonus plan (i.e., incentive vs. discretionary) such that firms
with relatively more assets-in-place are more likely to use incentive bonus plans.
If this is the case in general and other things are held the same, managers of these
firms are likely to have larger incentives to choose income-increasing accounting
procedures, implying an indirect link between the ios and accounting choice.
This is examined further in section 6.
For those firms with an incentive bonus plan that ties the bonus to an
accounting earnings measure (41 firms), I also examine whether there is a relation between how the parameters in these firms plans are set and their leverage
and ios measures. Of these 41 firms, 26 disclosed details of a lower bound and 16
disclosed details of an upper bound. The tests (not reported in tables) indicate
that there is no apparent relation between these bounds and the ios or leverage
of these firms. However, this is not too surprising since: (i) the results in table
3 indicate that firms with incentive bonus plans tend to be those with relatively
large amounts of assets-in-place, so there is not much variation in the ios of the
firms that disclose details of the upper and lower bounds, and (ii) the sample
sizes are small.
5.2.
I collect data
D. J. Skinner,
Incesrnzenr opporrunities
und accounting
choice
427
as seems likely. 2o Following Duke and Hunt (1990) and Press and Weintrop
(1990), I collect information on the existence of nine specific types of accountingbased covenants, although these nine covenants can be classified into the four
broad groups that both of these papers use: leverage restrictions, working
capital restrictions, net asset restrictions, and retained earnings restrictions.
Of the 504 sample firms, I am able to identify 465 in Moodys. Of these firms,
259 (56%) have at least one accounting-based debt covenant, while the remaining 206 do not. To examine the relation between accounting-based debt covenants, financial leverage, and the ios, table 4 provides a comparison of the firms
with at least one accounting-based
debt covenant to the firms without an
accounting-based debt covenant. 21 There are two principal results in the table.
First, consistent with the evidence in Begley (1990a) and Duke and Hunt (1990),
firms with accounting-based debt covenants have more debt than other firms.
The average (median) debt-to-assets ratio for the firms with at least one accounting-based debt covenant is 0.25 (0.23) compared to an average (median) of 0.16
(0.14) for the firms with no accounting-based debt covenants. These differences
are significant at the 1% level. Similar differences exist for the equity-to-value
leverage measure. Thus, firms with more debt are more likely to use accountingbased debt covenants in public debt agreements, as argued in section 2.
Second, there are also reliable differences between the ios measures for firms
with and without accounting-based debt covenants. On average, the firms with
accounting-based debt covenants in their public debt agreements have lower
asset betas (0.59 vs. 0.67), higher gross PP&E ratios (0.59 vs. 0.51), and smaller
R&D (0.018 vs. 0.028) and q (1.09 vs. 1.29) ratios than the firms without
accounting-based
covenants. Differences between the medians are of similar
magnitude, and all of these differences are statistically significant at the 5% level
for two-tailed, two-sample t and Wilcoxon tests. This evidence is consistent with
the prediction from section 2 that firms with more assets-in-place have more
debt, and so are more likely to use accounting-based debt covenants than other
firms [although again, this relation holds only for public debt - growth firms
2DAsdiscussed earlier, it is likely that managers in growth firms have more specific knowledge
about their firms assets and that the value of these growth opportunities is less observable. There is
therefore likely to be a relatively large informational asymmetry between managers in growth firms
and outsiders, especially if part of asset value comprises information that is proprietary. If this is so,
then these firms may favor the use of private debt over public debt for two reasons. First, contracting
and negotiation costs will be lower, because it is probably easier to convince a few relatively
sophisticated lenders of the value of their firms assets. Second, it is less likely that the proprietary
information will end up in the public domain if the firm negotiates with private lenders.
210f the firms with at least one accounting-based convenant, 69 had one covenant, 90 had two
covenants, 83 had three covenants, 14 had four covenants, and 3 had five covenants. When
I calculate the rank correlations between the number of accounting-based covenants and the
leverage and ios measures, I obtain similar inferences to those described in the text. That is, there is
a positive relation between the number of accounting-based debt covena?& financial leverage. and
assets-in-place.
D.J.
428
Skinner,
Inoestment
opportunities
and accounting
choice
Table 4
Comparison of average (median) size, leverage, and investment opportunity set measures for 465
firms with and without at least one accounting-based debt covenant in their public debt agreements
in fiscal 1987.eb
Firms with
accounting-based
debt covenants
(259)
Firm
Firms without
accounting-based
debt covenants
(206)
Test statistic
for
difference
size:
1,902
(656)
4,134
(616)
t = 2.84*
z = 0.21
Debt/assets
0.251
(0.227)
0.164
(0.138)
t = - 6.22*
Z = - 6.71*
MVE/value
0.521
(0.532)
0.654
(0.671)
t = 7.51*
Z = 6.99d
0.119
(0.123)
0.120
(0.126)
t = 0.22
z = 0.43
Asset beta
0.587
(0.597)
0.674
(0.663)
t = 2.71*
Z = 2.47*
PPE/value
0.588
(0.507)
0.508
(0.407)
t = - 1.95
Z = - 3.46*
R&D/sales
0.018
(0.002)
0.028
(0.007)
t = 2.95d
Z = 2.37d
Tobins q
1.09
(0.97)
1.29
(1.11)
t = 3.29d
z = 3.43*
Value
Finanrial
leverage:
Profitability:
Accounting ROA
Investment
opportunity
variables:
The numbers in the difference column are test statistics from two-sample t-tests (Wilcoxon tests)
of the null hypothesis that the average (median) difference between the two groups of firms is zero.
bThe variables are defined as follows. All variables (except asset beta, accounting ROA, and
Tobins q) are calculated for each firm as the average over year-end 1985 through year-end 1987, and
so are based on three observations for each firm.
Value
= market value of the firm (in millions of dollars)
market value of equity plus book value of debt;
Debt/assets 1 book value of long-term debt divided by the book value of total assets;
MVE/value = market value of equity divided by the market value of the firm;
Asset beta = beta of the firms stock multiplied by the MVE/value ratio;
PPE/value = gross property, plant, and equipment (at historic cost) divided by the market value
of the firm;
R&D/sales = research and development expense divided by net sales, expressed as a percentage;
Tobins q
= estimate of Tobins q ratio (see text for details of calculations).
Statistical significance at the 5% level (two-tailed test).
*Statistical significance at the I % level (two-tailed test).
429
may use more private debt which also uses accounting-based debt covenants
- see Leftwich (1983) 3. Thus, the evidence suggests that the ios drives both
financiai leverage and the nature of debt coniracts, which implies that there is
likely to be an indirect link between the ios and accounting choice, as outlined in
section 2. I address this directly next.
221 alsa use finer partition where suffkient detail is provided in the annual report. For example, if
a firm states that it uses a combination
oT the FIFO and AC methods, I code this choice as 1.5.
430
D.J. Skinner,
~~l~.~t~lle~ll
oppomrrilies
urtd
uCCouFltiFlg
CilOiCe
230f the 600 firms in the ATT sample, 559 use straight-line
depreciation,
132 use accelerated
methods, and 51 use units of production (clearly, a number of firms use more than one mclhod).
Similarly, of the 213 firms that report a goodwill amortization
period, 173 (HI%) choose 40 years
(and another 77 disclose that they use ;I period not exceeding 40 years).
D. J. Skitmer,
hw~rtmw
opporrutriries
unci accomtit~g
choice
431
Table 5
Summa ry of the inventory, depreciation. and goodwill accounting choices of 504 sample firms in
fiscal 1987.
_____
Choice
Code
Number
Percent of totaf
(A) Itmwtory
FIFO
AC,FIFO
AC or Mix
AC,LIFO
LIFO
cost jlow
assutnptiot~
2
1.5
1
0.5
0
152
26
97
15
178
Total
468
_______
(B) Depreciation
SL
UPiSL
UP or Mix
U P/Acc.
Act.
Total
__.__~~__
method
2
1.5
A.5
369
7
74
0
48
years
30-39 years
< 30 years
2
:,
40
Total
74.1
1.4
14.9
0.0
9.6
498
~~ _~~~~~_~ ______~_._
(C) Goodwill
attlortizatiot~
period
111
94
26
48.0
40.7
11.3
231
D~fitzition
FIFO
LIFO
AC
Mix
SL
Act.
UP
Mix
32.5
5.4
20.7
3.2
38.0
=
=
=
=
=
=
=
=
of acron~ws:
432
D.J. Skinner,
Incesstntenr opporlunities
and accounting
choice
while that on R&D is reliably positive, implying that firms with relatively more
assets-in-place are more likely to make the income-decreasing
accounting
choice.
Regression 4 in tab!e 6 includes firm size along with the four ios measures as
explanatory variables. Including the ios measures does not reduce the statistical
significance of the size relation, and conversely, the statistical significance of the
ios variables is not reduced by including firm size in regression. The p-value for
the joint null hypothesis that the coefficients on the ios variables are all zero is
significant at the 1% level and the pseudo R-square - of 14.3% - is larger than
for the regressions that include either just the size variable or just the ios
variables. It is again the case that the coefficient on the PP&E measure is
reliably negative while that on R&D is reliably positive. The inferences do not
change a great deal when I include all of the independent variables in regression
5. In addition to the size and ios variables, the ROA variable is negative and
significant, and the pseudo R-square increases to 15.4%. Overall, this evidence
indicates that the ios variables add explanatory power to the inventory regressions over and above that provided by firm size and ROA.
These results suggest that firms with relatively more assets-in-place are more
likely to choose LIFO, which I have characterized as the income-decreasing
accounting choice. However, the FIFO vs. LIFO choice also depends on these
firms relative tax positions and on the relative tax benefits of these two
accounting choices. The relative tax benefits of LIFO depend on the direction
and rate of change in firms input prices, on the variability of firms physical
inventory levels, and on firms effective tax rates [e.g., see Lee and Hsieh
(1985) and Dopuch and Pincus (1988)]. These characteristics are likely to be
correlated with the ios because: (i) the nature of the firms production process is
likely to affect the variability of inventory levels, and (ii) the firms production
process determines the inputs that are required. The more variable the firms
inventory levels, the higher the probability of forced LIFO liquidations, and SO
the less attractive is LIFO. Thus, firms with relatively more growth opportunities, because they also tend to be more risky, are less likely to select LIFO
than are other firms. So the tax effects of the firms ios generate a prediction
that firms with more assets-in-place will make the income-decreasing accounting choice.
To investigate the tax explanation, I include in regressions 6-10 in table
6 a variable (Infl.) that measures the direction and rate of price change in these
firms industries. Other things equal, the relative tax benefits of LIFO increase
with higher (more positive) rates of change in firms input prices. I use data from
the Bureau of Labor Statistics Producer Price Indices to group these firms
industries into three equal-sized groups: those with high, medium, and low rates
of price change since 1967 (similar to other studies, these are industry-level
output price indices). I then define the Infl. variable such that firms in the low
group are coded 2, those in the medium group are coded 1, and those in the high
0.36
(0.41)
- 0.26
( - 3.36)
10.
- 6.23
0.35
(1.60) ( - 2.02)
0.32
(1.50)
0.01
(0.02)
_~ ~~__
- 0.12
( - 0.25)
- 0.58
( - 1.19)
- 0.73
( - 1.54)
0.24
(0.80)
0.30
(0.99)
8.76
(2.17)
6.09
(1.44)
0.0800
0.0169
0.0000
0.0000
0.163
0.15 1
=
=
=
=
opportunity
The ios p-value is that from a test of the joint hypothesis that the coefficients on the ios proxy variables are all equal to zero. More specifically, the
ditference between minus twice the log-likelihood for the full model (including the ios variables) and minus twice the log-likelihood for the reduced model
(excluding the ios variables) follows an asymptotic chi-square distribution with degrees of freedom equal to the number of ios variables. See, e.g., Aldrich
and Nelson (1984, p. 59).
Beta
PPE
R&D
q
insestment
The table reports on multinomial logit regressions of firms inventory accounting scores (coded from 5 for FIFO through 1 for LIFO) on various
combinations of the following explanatory variables.
bAll explanatory variables (except beta, ROA. and the dummy variables) are calculated for each firm as the average over year-end 1985 through
year-end 1987 values, and so are based on three observations for each firm.
= market value of the firm (in millions of dollars)
Value
= market value of equity plus book value of debt;
Debtassets
= book value of long-term debt divided by the book value of total assets;
D:A*covenant = debt,assets multiplied by a dummy variable set equal to one if the firm has at least one accounting-based debt covenant and zero
otherwise;
= dummy variable set to one if the firm has a bonus plan where the amount of the award is explicitly tied to an accounting earnings
Bonus
number and zero otherwise:
= ten-year average of the yearly ratio of operating income (before depreciation expense and after adjusting for the change in the LIFO
ROA
reserve) to firm value. expressed as a percentage;
= variable coded as 2/1/O for firms in industries that experienced low/medium/high rates of price change since 1967, as measured by the
Infl.
Bureau of Labor Statistics Producrr Price Indices.
0.47
(0.53)
- 0.31
( - 4.07)
9.
436
D.J. Skinner,
Jncestment
opportunities
and accounting
choice
group are coded 0, If the relative tax benefits of LIFO increase with higher rates
of increase in firms input prices (and this helps explain the accounting choice),
the Infl. variable will be positively related to accounting choice.25
The evidence indicates that tax-related incentives do help explain firms
inventory method choices. When I include the Ink variable with firm size and
financial leverage in regression 6, its coefficient is reliably positive (t-statistic of
3.77), consistent with the prediction that the relative tax benefits of LIFO
increase with increases in input prices. The coefficient on firm size remains
reliably negative. Similarly, when I include the Infl. variable along with firm size,
financial leverage, and the ROA measure in regression 7, the coefficients on firm
size and ROA are reliably negative and that of InIX reliably positive.26 [Some
care must be exercised in comparing the results in equations 6-10 with those in
equations l-5 since there are 357 firms with available data for equations l-5 but
only 256 firms with available data for equations 6-10 (i.e., there are 101 firms
without the inflation data).]
Regressions 9 and 10 in table 6 include the ios variables along with firm size,
ROA, and the InP.. measure. The ios measures are less significant when all of
these other variables are included than they were in regressions l-5, suggesting
that the ios is correlated with tax incentives. (The fact that the ios variables are
jointly significant at the 0.0002 level in regression 8 indicates that the different
sample is not the cause of the decline in significance for these variables.)
Consistent with this, firms with relatively more growth opportunities tend to be
those for which the tax benefits of LIFO are relatively small: the InIl.variable is
negatively related to the PP&E variable (correlation of - 0.42) and positively
related to R&D (0.25) and 4 (0.25) so that firms with relatively more assetsin-place tend to have experienced larger rates of price increase.
2JI have also used this variable to perform a check on my coding of LIFO as income-decreasing
and FIFO as income-increasing. If input prices are falling over some period of time, then FIFO is
likely to be the income-decreasing accounting choice over that period. Therefore, it may be that my
coding is incorrect for firms for which the price variable is coded 2 (relatively low rates of price
increase) and correct for firms for which this variable is coded 0 (relatively high rates of price
increase). If this is the case, I would expect a different relation between the probability of choosing
FIFO and the independent variables for different levels of the price variable. However, when I run
these regressions with the price variable included as an interactive dummy with the other independent variables, none of the interactive terms are significantly different from zero at the 10% level,
which provides some assurance about the way that I code the inventory choice variable (these
regressions are not reported in tables),
261n unreported regressions I also include a measure ot :hese firms net operating loss (NOL)
carryforwards. Firms with relatively large NOL carryforwartis have little incentive to choose LIFO
to minimize taxes. Consistent with this argument, I find that firms that use FIFO have significantly
larger NOL carryforwards than firms that use LIFO (the mean and median differences are
significant at the 5% level). However, the NOL carryforward variable is not significant in a multivariate logit that also includes firm size, ROA, and lnfl. as independent variables. This could be due
to collinearity: not surprisingly, the NOL carryforward variable is negatively related to ROA
(correlation of - 0.44).
accounting choice
437
Panel A of table 7 reports ehe results for the depreciation method choice. The
first regression in panel A characterizes extant results for the size, debt/equity,
and bonus plan hypotheses. Consistent with these hypotheses, the probability of
choosing the income-increasing accounting alternative (in this case straight-line
depreciation) increases as financial leverage increases, is larger when the firm has
a bonus plan that ties the bonus directly to accounting earnings, and decreases
as firm size increases. The coefficients on all three variables are reliably different
from zero, with t-statistics of - 3.72 (firm size), 4.09 (financial leverage), and 3.32
(bonus plan), and the pseudo R-square for this regression is 9.1 percent. The
second regression includes the ROA measure along with the size, leverage, and
bonus play variables. The coefficient on ROA is. insignificant.
According to the debt/equity hypothesis, financial leverage is important in
these regressions because it proxies for managers incentives to choose incomeincreasing accounting procedures to loosen accounting-based debt covenants. If
this is the case, there will be a relation between accounting choice and financial
leverage for firms with accounting-based covenants but not for firms without
such covenants. To test this, in regression 3 I include a dummy variable coded
1 if the firm has at least one accounting-based debt covenant and 0 otherwise,
multiplied by the debt/assets variable, as well as including the debt/assets
variable alone. The coefficient on this multiplicative dummy variable measures
how the slope coefficient on debt/assets differs for firms with and without
accounting-based debt covenants.
Similar to the results in Press and Weintrop (1990), the leverage variable
remains significant (t-statistic of 2.46) in this regression while the multiplicative
dummy variable is insignificant (t = 0.74). Thus, the relation between leverage
and the depreciation choice does not seem to be due to accounting-based
covenants, and so may reflect a direct effect of the ios on both accounting choice
and leverage. (However, this test is probably biased against the multiplicative
dummy variable because the debt/assets variable includes private as well as
public debt, while the debt covenant dummy includes only covenants in public
debt agreements.)
Regression 4 in panel A of table 7 includes the four ios measures. In this
regression the null hypothesis that all of the coefficients on the ios proxy
variables are zero is rejected at the 7% level. However, the only variable that is
individually significant is asset beta, which has a negative coefficient, implying
that firms with higher betas are more likely to select income-decreasing accounting procedures.
When I include the ios measures along with all of the other variables in
regression 5, they are only jointiy significant at the 10% level, and none of the
ios variables is individually significant. In contrast, the firm size, financial
leverage, and bonus plan variables remain statistically significant (their
t-statistics are largely unaffected by including the ios measures). Moreover, the
pseudo R-square in this regression is 9.8% which is similar to that for the first
438
opportunities
and accounting
choice
regression in the table. The traditional variables thus seem to be more important than the ios variables. Because of collinearity in these data, this result does
not necessarily imply that size, leverage, and the bonus plan are the right
variables and that the ios measures are not. However, the fact that the
size, leverage, and bonus plan relations that other studies document are robust
to including the ios variables is important, because it suggests that these
variables do not matter in other studies merely because they are correlated
with the ios.
Table 7, panel B provides results for the choice of goodwill amortization
period. The results in the first regression are similar to, but weaker than, those
for depreciation: accounting choice is negatively related to firm size and positively related to both debt/assets and the existence of a bonus plan. However,
only the coefficient on debt/assets is reliably different from zero. The different
results may be because tlic sample of firms here is different to those in the
depreciation and inventory regressions: only 171 firms are included in panel
B compared with 369 firms in panel A and 357 firms in table 6. Moreover, the
171 firms included here tend to be larger than those in the depreciation
regressions (the difference in medians is significant at the 1% level) because the
smaller sample firms tend not to report a goodwill choice. When I reestimate the
depreciation regressions with only the 169 firms that report both goodwill and
depreciation choices, the size and bonus plan relations become insignificant.
In regression 2 of panel B I again include the multiplicative dummy variable
to see whether it is leverage per se,or the existence of accounting-based debt
covenants in combination with leverage, that better explains accounting choice.
In this regression the coefficient on the multiplicative dummy variable is positive
and significant at the 10% level, while the coefficient on debt/assets becomes
insignificant. Thus, for these firms leverage is only important for firms with
accounting-based covenants, which is consistent with managers selecting income-increasing accounting procedures to loosen debt-covenant constraints.27
Regression 3 includes the four ios measures as explanatory variabies. The ios
p-value is small (less than O.OOl), indicating that the ios variables together are
related to the goodwill choice, and the pseudo R-square is 10.2%, larger than for
either of the first two regressions in panel B. The two variables that are most
significant are PP&E (t-statistic of - 1.75) and R&D (t-statistic of - 3.45).
While the negative coefficient on R&D suggests that firms with more growth
opportunities are more likely to select income-decreasing accounting procedures, the negative coefficient on gross PP&E suggests the opposite.
*These results are different to those in panel A: the multiplicative dummy variable was not
significant in the depreciation regressions, while the debt/assets variable was significant. This
difference is not due to the different samples - when I reestimate the depreciation regressions with
only the 171 firms included in panel B, the coefficient on the debt/assets variable is still reliably
positive, while that on the multiplicative dummy is still insignificant.
D. J. Skinner, hcesmrenr
439
1 do not include the inventory choice because, as noted above, this choice depends on tax as well
as contracting considerations, and so seems to be a different decision from the other two accounting
choices.
29Since the goodwill choice has only three levels (income-increasing, intermediate, and incomedecreasing), I also collapse the depreciation choice into three levels. Since there are three levels for
each of two choices, there are nine possible permutations and five unique strategies. Of the 229 firms
that report both choices, 85 (37%) select strategy 5 (the most income-increasing strategy), 85 (37%)
select strategy 4,39 (17%) select strategy 3, 14 (6%) select strategy 2, and 6 (3%) select strategy 1 (the
most income-decreasing strategy).
- 0.18
- 0.10
- 0.17
( - 0.10)
- 0.03
( - 0.32)
( - 0.22)
5.
- 0.35
0.70
(0.49)
2.47
(2.35)
- 0.02
( - 0.23)
( - 1.05)
- 0.09
( - 1.16)
4.
3.
7
-.
1.
3.46
(2.55)
3.04
(2.14)
2.64
(1.93)
2.49
(1.87)
1.17
(0.96)
0.83
(3.19)
_.
3.83
(1.20)
1.99
(0.69)
0.43
0.56
(1.69)
0.62
(1.88)
( 1.64)
0.51
(1.40)
PP&E
( - 0.61)
- 0.18
- 0.15
( - 0.54)
R&D
- 17.25
(- 3.16)
0.05
- 1.15
(0.07) ( - 1.56)
- 5.25
( - 1.19)
- 17.45
( - 3.45)
- 18.83
( - 3.52)
- 1.17
( - 1.75)
2.81
(0.74)
- 7.87
( - 1.93)
- 0.56
( - 1.03)
- 0.31
( - 0.09)
( - 0.97)
- 0.44
( - 3.52)
- 0.26
0.88
(0.74)
0.83
(3.23)
0.85
(3.32)
Beta
____-
0.0001
0.0004
0.51
(i-81)
0.32
(1.04)
0.0007
0.0976
0.28
(0.96)
0.19
(0.74)
0.0703
ios
p-value
4
___~~__
- 0.02
( - 0.09)
10s meusures:
___~~__ __-
ROA
ROA:
5.
( - 2.97)
3.10
(2.46)
4.04
(4. IO)
- 0.25
( - 3.74)
4.05
(4.09)
- 0.23
( - 3.72)
Bonus
Bonus plan:
___
- 1.05
( - 2.67)
3.
7
_.
1.
D/A *
covenant
Leuerqe:
__-_-
Debt/assets
____
____
Value
~--
Size:
4.
0.0003
0.0001
0.0007
0.0091
0.0186
0.0000
0.0703
0.153
0.146
0.102
0.073
0.055
0.098
0.023
0.067
0.065
o.oooo
o.oooo
0.091
Pseudo
R-square
0.0000
Overall
model
p-value
Estimated coefficients (with asymptotic t-statistics in parentheses) from a multinomial logit model of the depreciation (panel A), goodwill (panel B), and
combined depreciation-goodwill
method choices of sample firms in fiscal 1987, estimation by maximum likelihood. A positive coefficient indicates
a higher probability of choosing a more income-increasing accounting method.a*b
Table 7
2.03
(1.19)
2.69
(1.85)
4.19
(3.86)
2.42
(1.75)
2.02
(1.53)
0.95
(2.97)
0.94
(3.06)
0.88
(2.87)
- I.17
( - 0.28)
- 2.69
( - 0.74)
- 0.40
( - 0.66)
- 1.11
( - 2.05)
- 1.23
( - 2.38)
- 0.98
( - 1.40)
- 0.63
( - 0.96)
- 0.75
( - 1.19)
0.31
(1.06)
0.18
(0.66)
- 13.9
( - 2.87)
- 10.4
( - 2.06)
0.18
(0.65)
- 13.2
( - 2.82)
0.0012
0.0001
0.0008
0.0000
0.0020
0.192
0.101
0.100
0.154
o.oooo
0.0008
0.14s
o.oooo
The ios p-value is that from a test of the joint hypothesis that the coefficients on the ios proxy variables are all equal to zero. More specifically, the
difference between minus twice the log-likelihood
for the full model (including the ios variables) and minus twice the log-likelihood
for the reduced model
(excluding the ios variables) follows an asymptotic chi-square distribution
with degrees of freedom equal to the number of ios variables. See, e.g., Aldrich
and Nelson (1984, p. 59).
Investment
=
Beta
PPE
=
R&D =
=
4
5.
4.
- 0.14
( - 1.58)
- 0.16
( - 1.92)
3
__
3.
- 0.17
( - 2.02)
1.
442
coefficients on R&D and beta are negative and reliably different from zero,
again suggesting that firms with more assets-in-place (fewer growth opportunities) are more likely to make the income-increasing accounting choice. In
regression 4 I include the ROA variable along with the ios variables. Similar to
the results in panels A and B, the ROA variable is not significant, and its
inclusion does not affect the significance of the ios variables.
When I include all of the variables in regression 5, the pseudo R-square
increases to 19.2% but only two of the coefficients are individually significant:
the coefficient on the bonus plan dummy is reliably positive and the coefficient
on R&D is reliably negative. Of the other variables, the one that is closest to
being significant at the 5% level is that on the multiplicative debt/assets
*covenants variable. The reduction in the individual significance of the other
variables probably reflects the collinearity that exists between these variables.
Overall, the evidence in table 7 suggests that the traditional explanations are
important, even after controlling for the effect of the ios. Nevertheless, the ios
variables do help explain these choices, both by themselves and in combination
with the traditional variables. The results in table 7 also suggest that firms with
more assets-in-place are more likely to choose income-increasing accounting
procedures, which is opposite to the result for the inventory choice.
443
Third, the ios variables are correlated with all three of the accounting
procedures that I examine. Moreover, except in the case of depreciation, the ios
variables provide incremental explanatory power in regressions that also include
firm size, financial leverage, and the bonus plan dummy as independent variables. Thus, the evidence indicates that there is an association between the ios
and accounting procedure choice, even after controlling for managers contractual incentives to select particular accounting procedures.
More specifically, firms with larger ratios of assets-in-place to value are more
likely to make income-increasing depreciation and goodwill choices than other
firms. However, while the ios proxy variables are also related to these firms
inventory accounting choices, the directional association is opposite that for the
depreciation and goodwill choices: firms with larger ratios of assets-in-place to
value are more likely to make the income-decreasing accounting choice (LIFO)
than are other firms. One plausible explanation for this result is that firms
inventory method choices are also affected by the relative tax costs of LIFO vs.
FIFO, and that the relative tax costs of FIFO are smaller for firms with
relatively more growth opportunities.
Overall, the evidence in this paper increases our understanding of current
accounting practice in several ways. First, while there has been speculation in
the literature about the effect of the ios on accounting choice, this study is the
first to document this relation in a general way for a large, randomly-chosen,
sample of firms. Importantly, the evidence suggests that previous evidence on
the size, debt/equity, and bonus plan hypotheses cannot be explained by the fact
that those studies exclude the ios, which is a potentially important correlated
omitted variable. Second, this study documents that the ios systematically
affects the nature of firm contracts, and thus affects managers contractual
incentives to select particular accounting procedures. Finally, the evidence
provides some limited support for the view that the ios affects accounting choice
directly, which is expected if the ios affects which accounting procedures are
optimal ex ante.
References
AICPA, 1988, Accounting trends and techniques, 42nd ed. (AICPA. New York, NY).
Aldrich, John H. and Forrest D. Nelson, 1984, Linear probability, logit, and probit models (Sage
Publications, Beverly Hills, CA).
Ball, Ray and George Foster, 1982, Corporate financial reporting: A methodological review of
empirical research, Supplement to Journal of Accounting Research 20, 161-234.
Begley, Joy, 1990a, The use of debt covenants to control agency problems, Unpublished dissertation
(University of Rochester, Rochester, NY).
Begley, Joy, 1990b, Debt covenants and accounting choice, Journal of Accounting and Economics
12, 125-139.
Bradley, Michael, Gregg A. Jarrell, and E. Han Kim, 1984 On the existence of an optimal capital
structure: Theory and evidence, Journal of Finance 39, 8:7-880.
Clinch, Greg, 1991, Employee compensatton a,rcf fins resear-.-h and development activity, Journal
of Accounting Research 29. 59-78.
444
opprtunities
and accounting
choice
Collins, Daniel W.. S.P. Kothari, Jay Shanken, and Richard G. Sloan, i992, Lack of !imeliness
versus noise as explanations for low contemporaneous returns<arnings relation, Working paper
(University of Rochester, Rochester, NY).
COX,
D.R. and E.J. Snell, 1989, Analysis of binary data, 2nd ed. (Chapman and Hall, London).
Dechow, Patricia M. and Richard G. Sloan, 1991, Executive incentives and the horizon problem: An
empirical investigation, Journal of Accounting and Economics 14, 51-89.
Demski, Joel S., James M. Patell, and Mark A. Wolfson, 1984, Decentralized choice of monitoring
systems, The Accounting Review 59, 16-34.
Dhaliwal, Dan S., Gerald Salamon, and E. Dan Smith, 1982, The effect of owner versus management
control on the choice of accounting methods, Journal of Accounting and Economics 4,41-53.
Dopuch, Nicholas and Morton Pincus, 1988, Evidence on the choice of inventory accounting
methods: LIFO versus FIFO, Journal of Accounting Research 26, 28-59.
Duke, Joanne C. and Herbert G. Hunt III, 1990, An empirical examination of debt covenant
restrictions and accounting-related debt proxies, Journal of Accounting and Economics 12,
45-63.
Foster, George, 1986, Financial statement analysis. 2nd ed. (Prentice-Hall, Englewood Cliffs, NJ).
Gaver, Jennifer J. and Kenneth M. Gaver, 1993, Additional evidence on the association between the
investment opportunity set and corporate financing, dividend and compensation policies, Journal of Accounting and Economics 16, forthcoming.
Hagerman, Robert L. and Mark E. Zmijewski, 1979, Some economic determinants of accounting
policy choice, Journal of Accounting and Economics 1, 141-161.
Hall, Bronwyn H., Clint Cummins, Elizabeth S. Laderman, and Joy Mundy, 1988, The R&D master
file documentation, Technical working paper no. 72 (National Bureau of Economic Research,
Cambridge, MA).
Hamada, Robert S., 1972, The effects of the firms capital structure on the systematic risk of common
stocks, Jottrnal of Finance 27,435452.
Healy, Paul M., 1985, The effect of bonus schemes on accounting decisions, Journal of Accounting
and Economics 7,85-107.
Holthausen, Robert W., 1990, Accounting method choice: Opportunistic behavior, efficient contracting, and information perspectives, Journal of Accounting and Economics 12, 207-218.
Holthausen, Robert W. and Richard W. Leftwich, 1983, The economic consequences of accounting
choice: Implications of costly contracting and monitoring, Journal of Accounting and Economics 5, 77- 117.
Kim, Kyungho, and Douglas A. Schroeder, i990, Analysts use of managerial bonus incentives in
forecasting earnings, Journal of Accounting and Economics 13. 3-23.
Lambert, Richard A. and David F. Larcker, 1987, An analysis of the use of accounting ;:nd market
measures of performance in executive compensation contracts, Journal of Accounting Research
25,85-125.
Lee, Chi-Wen Jevons and David A. Hsieh, 1985, Choice of inventory accounting meth,ods: Comparative analyses of alternative hypotheses, Journal of Accounting Research 23, 468-485.
Leftwich, Richard, 1983, Accounting information in private markets: Evidence from private lending
agreements, The Accounting Review 58, 23-42.
Leftwich, Richard, 1990, Aggregation of test statistics: Statistics vs. economics, Journal of Accounting and Economics 12, 37-44.
Lilien, Steven, Martin Mellman, and Victor Pastena, 1988, Accounting changes: Successful versus
unsuccessful firms, The Accounting Review 63, 642-656.
Lindenberg, Eric B. and Stephen A. Ross, 1981, Tobins r~ratio and industrial organization, Journal
of Business 54, l-32.
Long, Michael S. and Ileen B. Malitz, 1985, Investment patterns and financial leverage, in: B.
Friedman, ed., Corporate capital structure in the United States (University of Chicago Press,
Chicago, IL) 325-348.
Maddala, G.S., 1983, Limited-dependent and qualitative variables in econometrics, Econometrics
Society monograph (Cambridge University Press, Cambridge).
Merck, Randall, Andrei Shleifer, and Robert W. Vishny, 1988, Management ownership and market
valuation: An empirical analysis, Journal of Financial Economics 20, 293-315.
Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147-175.
445
Press, Eric G. and Joseph B. Wcinrrop, ! 390, Accounting-based constraints in public and private
debt agreements: Their association with leverage and their
I, impact on accounting choice, Journal
of Accounting and Economics 12,65-95.
Sibson, Koberi E., i99g, Compensation, 5th ed. (American Management Association, New York,
NY)
Smith, Clifford W., Jr. and Jerold B. Warner, 1979, On financial contrac;ir~~*
.e. An analysis of bond
covenants, Journal of Financial Economics 7, 117-161.
Smith, Clifford W., Jr. and Ross L. Watts, 1982, Incentive and tax effects of executive compensation
plans, Australian Journal of Management 7, 139-157.
Smith, Clifford W., Jr. and Ross L. Watts, 1991, The investment opportunity set and corporate
policy choices, Working paper (University of Rochester, Rochester, NY).
Sweeney, Amy P., 1991, The impact of debt-covenant violations on accounting choices, Working
paper (Harvard Business School, Cambridge, MA).
Watts, Ross L.. 1977, Corporate financial statements, a product of the market and political
processes, Australian Journal of Management 2, 53-78.
Watts, Ross L. and Jerold L. Zimmermman, 1986, Positive accounting theory (Prentice-Hall,
Englewood Cliffs, NJ).
Watts, Ross L. and Jerold L. Zimmerman, 1990, Positive accounting theory: A ten year perspective,
The Accounting Review 65, 131-156.
Zimmer, Ian, 1986, Accounting for interest by real estate developers. Journal of Accounting and
Economics 8, 37-5 1.
Zmijewski, Mark E. and Robert L. Hagerman, 1981, An income strategy approach to the positive
theory of accounting standard setting/choice, Journal of Accounting and Economics 3,129-149.